[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date: Mon, 8 Aug 2016 11:45:58 -0700
From: Linus Torvalds <torvalds@...ux-foundation.org>
To: Borislav Petkov <bp@...en8.de>
Cc: Ville Syrjälä <ville.syrjala@...ux.intel.com>,
Linux Kernel Mailing List <linux-kernel@...r.kernel.org>,
"the arch/x86 maintainers" <x86@...nel.org>,
Borislav Petkov <bp@...e.de>,
"H . Peter Anvin" <hpa@...or.com>,
Andy Lutomirski <luto@...capital.net>,
Brian Gerst <brgerst@...il.com>,
Denys Vlasenko <dvlasenk@...hat.com>,
Peter Zijlstra <peterz@...radead.org>,
Thomas Gleixner <tglx@...utronix.de>,
Ingo Molnar <mingo@...nel.org>, Jiri Kosina <jkosina@...e.cz>
Subject: Re: [PATCH] x86/hweight: Don't clobber %rdi
On Mon, Aug 8, 2016 at 11:37 AM, Borislav Petkov <bp@...en8.de> wrote:
>
> Ok, so do you think it would work too if I stated that the input
> register gets clobbered:
>
> asm (ALTERNATIVE("call __sw_hweight64", POPCNT64, X86_FEATURE_POPCNT)
> : "="REG_OUT (res)
> : REG_IN (w)
> : REG_IN);
That would work (although the clobbered registers have a different
syntax than the in/out registers), but it would be wrong, in my
opinion.
We want the actual POPCNT instruction to be the common case, and that
instruction does *not* clobber any other registers than the output.
So I think it's much better to just say: "the __sw_hweight functions
should have the same semantics as popcnt" (although without the eflags
rules that we don't care about).
There's nothing wrong with keeping it as assembly language - it's not
like it's a maintenance headache once it is written.
Linus
Powered by blists - more mailing lists